Seed Media Group: Blog
Saturday, December 19, 2009 • by Eva Wisten • #
Some reflective common sense thoughts
Bo Morgan from MIT Media Lab came by this week and explained how to
make little artificial Carol learn from her mistakes.
Bo Morgan, a PhD candidate at the Media Lab, is exploring the
old but small field of reflective common sense thinking: how
to reflect, learn from mistakes and avoid them in order to accomplish a
goal. By using the models presented in the Emotion Machine
written by his advisor, AI pioneer Marvin Minsky, Morgan is
trying to “make little Carol who is playing in a sand box figure out
how to fill her cup with mud”.
Morgan and his group are trying to emulate human mental execution by
steps like dividing the cognitive process into six layers, assigning
the program a goal, give it the capability to draw from memory and to
take on different identities depending on the situation. By coding
according to theories on reflective thinking, Morgan hopes to gain
insights to how we can create AI-tools that debug themselves as they
run and create robot reasoning that is as robust as our own.
“I’m programming a model that’s relatable to evolution. Algorithms
that solve problems are rewarded with survival,” Morgan says.
Morgan’s dissertation is expected in September next year. After that
he hopes to apply his knowledge to the problems of mental illness and
cognitive dysfunction by performing neuropsychological experiments to
empirically test his theories.
